Research on pre-task planning to date has mainly focused on task performance. However, the effects of planning are contingent on what learners actually do during planning time. One important factor that may determine the quality and usefulness of planning is whether it is done in the first language (L1) or the second language (L2). This research addresses this issue by investigating the relative benefits of collaborative planning in the L1 and L2 in terms of ideas generated and transferred to an oral problem-solving task. Seventy-two Japanese university EFL learners were randomly assigned to one of two planning conditions: L1P (L1 planning, Japanese) and L2P (L2 planning, English). Dyads in each group were given 10 minutes to plan the content of a problem-solving task in the respective languages before individually performing the timed 2.5-minute oral task. Data took the form of transcribed planning discussions and transcribed task performances. All data were coded for idea units and sorted into categories of problem–solution discourse structure (situation, problem, response, evaluation). A qualitative comparison of L1 and L2 planners’ generation of idea units during planning, transfer and performance was conducted to supplement the quantitative analysis. Findings indicate the L1P condition has significant advantages over the L2P condition in terms of idea conceptualization, but this advantage had a limited impact on subsequent L2 task performance. Pedagogical implications are discussed in terms of possibilities for productively incorporating L1 planning during task implementation in foreign language contexts where learners share a common first language.